Energy prices have risen dramatically in recent years. Additionally, the public has become increasingly aware of the environmental impact of large-scale energy production and distribution. Both companies and individuals are looking for solutions to address escalating energy costs, as well as mitigating environmental impact through reduced energy consumption.
Heating, cooling, and lighting of residential and commercial buildings are major elements of overall energy consumption within the economy, and significant contributors to the energy costs for individuals and businesses. Energy required for residential and commercial buildings can be reduced through behavior modification by, for example, reducing heating and cooling during off-hours and turning off lights when not needed. However, the problem with behavior modification is that, even with the best intentions, people find it difficult to change their habits.
Governmental and environmental organizations have suggested, and in some cases even mandated, that the public use programmable energy saving devices, such as programmable setback thermostats, to reduce energy consumption. However, these devices have a minimal beneficial effect on energy savings because they are either improperly programmed or are not programmed at all.
To date, methods that attempt to address the problem fall into one of two basic categories, reactive and predictive. Reactive methods drive an output in reaction to receiving an input, such as a signal from an artificial light sensor or motion sensor. During periods lacking input, reactive methods, often after a time delay, enter into a setback or energy saving mode. Upon receiving input, these methods revert to normal operation. Predictive methods, on the other hand, gather and tabulate input and attempt to schedule output changes according to a derived schedule. Reactive methods can lead to occupant discomfort by driving rapid output change in response to input while predictive methods can lead to mistimed output due to faulty prediction.
The reactive method described by Riley (U.S. Pat. No. 5,395,042) attempts to mitigate the severity of reactive output, in this case heating and cooling, by controlling tolerances in order to minimize recovery time. For instance, the system can be programmed to allow a temperature drift no greater than that which can be restored within a given time period (e.g., 15 minutes).
The predictive method described by Day-Theuer (U.S. Pat. No. 6,375,087) yields poor scheduling results. In particular, the apparatus of Day-Theuer can be easily confused by sporadic sensor events outside the norm because it correlates all sensor events equally
The predictive method described by Bell (U.S. Pat. No. 5,088,645) attempts to solve the problem of unusual occurrences by linking output levels directly to the frequency of time-aligned historical inputs. However, this method has very undesirable consequences due to the association between input frequency and output level. Sporadic input by sensors located in low traffic areas, hallways for instance, will produce instability in the method and lead to severe output oscillations.
Bodmer (U.S. Pat. No. 6,263,260) describes a method relying on neural networks to perform a stochastic analysis of sensor data in order to generate better predictive output. Unfortunately, neural networks require tremendous parallel computing power and significant memory resources. This requirement is counterproductive, as more energy may be required to implement the method than the amount of energy saved.